driver mesh

[joevenzon]

OK, I put the driver mesh in VDrift R1979 (data R108). The driver's position is controlled by the driver position value in the .car file. Most of the car files need to be tweaked to position the driver a little better.

[zimluura]

i'm working on making the driver look a little bit better. tweaking the mesh, adding a few more faces, and increasing the texture resolution.

but he'll still be keeping the same dimensions.

i think the .blend file was in the zip, if you want to leave that in the driver's data folder so people can have a .blend file to start from with additional drivers it's a-o-k by me.

could be useful as a basic humanoid to start tweaking
or just to get a solid idea for positioning an entirely new mesh.
the .blend could also be useful for making skins.

[joevenzon]

OK, I added the .blend file to the art repository.

[zimluura]

got the latest source and compiled it, also got all the newest data. i'm really happy with the changes. the whole thing doesn't seem to want to crash, and the shiny car bodies look sweet now!!

i can go ahead and work on updating all the cars so they look right with the driver. it's just grunt work really.

but i don't know much about using svn. i would guess i can just commit the slightly modified .car files with tortoise svn but i don't see an obvious way to do that right now. i'll probably also need increased rights with the server or something, but if you want me to be able to do that make sure that i won't be able to permanently harm anything by being an svn noob. and if you don't want me to have that direct a relationship with the svn server then that's cool too.

// here are some random thoughts from being up all night
should the driver be rendered on an interior camera? and if that's going to stay; should i just try to put the camera right in front of his eyes?

the ff doesn't quite work with him (i think no matter how much positioning is done his arms and legs will bulge out of the car. not really a big deal imo. but i could make an alternate mesh with higher legs and slimmer shoulders. same skin and all though, same driver folder i guess, call it body_compact.joe. then maybe car makers can specify if their car needs the compace mesh...sorta adds overhead to the driver making pipeline though, but not really allot.

i've been working on refining the driver model higher poly (1310 total) + more detailed higher resolution skin. since this driver mesh will never be animated i might try to re do the skin mapping so front and back are one big section and the connecting seams are like this: one down the center of his back forking out across his shoulder blades, down the back of each leg, and down each arm on its underside (like armpit to elbow to bottom edge of hand). thinking about all the places you're least likely to easily see.

any thoughts on any of this?

[joevenzon]

i can go ahead and work on updating all the cars so they look right with the driver. it's just grunt work really.

Yeah, sure, if you don't mind!

but i don't know much about using svn. i would guess i can just commit the slightly modified .car files with tortoise svn but i don't see an obvious way to do that right now. i'll probably also need increased rights with the server or something, but if you want me to be able to do that make sure that i won't be able to permanently harm anything by being an svn noob. and if you don't want me to have that direct a relationship with the svn server then that's cool too.

Sure, I can set you up with SVN commit access to the data repository. What's your sourceforge.net username?

Tortoise SVN documentation is here:
http://tortoisesvn.net/docs/release/Tort...index.html

But basically you can just right click on your data/cars folder and do a check in and it'll automatically check in your changes to files you modified under that folder. Just make sure you don't check in extra things that you don't really mean to check in.

[/quote]should the driver be rendered on an interior camera? and if that's going to stay; should i just try to put the camera right in front of his eyes?[/quote]

Yeah, I think so.

the ff doesn't quite work with him (i think no matter how much positioning is done his arms and legs will bulge out of the car. not really a big deal imo. but i could make an alternate mesh with higher legs and slimmer shoulders. same skin and all though, same driver folder i guess, call it body_compact.joe. then maybe car makers can specify if their car needs the compace mesh...sorta adds overhead to the driver making pipeline though, but not really allot.

Ah, don't worry about it for right now. I think eventually there will need to be a skeletal animation system that allows positioning of the arms in a per-car way anyway.

i've been working on refining the driver model higher poly (1310 total) + more detailed higher resolution skin. since this driver mesh will never be animated i might try to re do the skin mapping so front and back are one big section and the connecting seams are like this: one down the center of his back forking out across his shoulder blades, down the back of each leg, and down each arm on its underside (like armpit to elbow to bottom edge of hand). thinking about all the places you're least likely to easily see.

Hmm, I didn't quite follow this.

[zimluura]

that last bit was about a technique for wrapping the skin around the driver. have all the seams where they're really hard to see so that skin making is super easy and you don't need to make sure the pixels on one either side of the seam line up properly.

at first i tried to make them like clothing seams thinking i could just darken them, but it doesn't look as good as it could were you allowed to do complex per pixel clothing seams.

wondering if anyone had any input on that idea for doing the seams of the driver (i.e.; trying to hide them instead of use them). 'cause i'm not a very experienced skinner.

====
on the sourceforge thing. i thought it would be this zimluura account, but that might just be for the forums here. not sure i have a sourceforge account then. tried for a second to access sourceforge.net, but it's down at 4:02am edt i'll try to look into that over the weekend.

[joevenzon]

wondering if anyone had any input on that idea for doing the seams of the driver (i.e.; trying to hide them instead of use them). 'cause i'm not a very experienced skinner.

Ohh, okay, I understand now. No input from me; I can't skin things worth a damn. :-)

on the sourceforge thing. i thought it would be this zimluura account, but that might just be for the forums here. not sure i have a sourceforge account then. tried for a second to access sourceforge.net, but it's down at 4:02am edt i'll try to look into that over the weekend.

OK. I'm gone this weekend but just post about it and I'll add you next week.

[zimluura]

wow, looks like i setup a sourceforge account in the beginning of 2003 and totally forgot about it. but anyway, i can login to it. so it's good.

zimluura
user id: 681661

i wonder what i'd intended to do with it then.

Edit - forgot to mention something: I did all the driver mass positions but...it looks like, right now the view position of the driver uses a different coordinate system than the mass positioning.

* mass positioning
car x = right & left, negative being left
car y = forward & back, negative being back
car z = up & down, negative being down

* view positioning
view x = forward & back, negative being back
view y = right & left, negative being right
view z = up and down, negative being down

is this desired behavior? if not would it be easy to straighten out before i re-do all the view positions in the .car files?

[joevenzon]

OK, I added you to the project on sourceforge so you should have SVN access to the data repository.

So, the coordinate systems inside VDrift are a mess. The intent is for all of the .car file coordinate systems in sync. Can you test what the correct coordinate system is (look at wheel positioning for the "right" coordinate system) and then see how the view and driver mass positioning differs?

[zimluura]

hmm, well i already had the mass and view positions. but i'll copy them to here as well.

* wheel positioning (tested on Front Right wheel)
wheel x = forward and back, negative being back
wheel y = left and right, negative being right
wheel z = up and down, negative being down

* mass positioning (same as blender)
car x = right & left, negative being left
car y = forward & back, negative being back
car z = up & down, negative being down

* view positioning (same as wheel)
view x = forward & back, negative being back
view y = right & left, negative being right
view z = up and down, negative being down

imo, mass seems like the system to use. it's how you'd draw a car from overhead with it's front pointing to the top of the paper. it also seems to be correct for blender importing.

wheel positioning looks sorta messed up to me. exactly the same way the view coordinates are.

[zimluura]

committed a much nicer looking driver model

[reece146]

Random thought of the day...




Who's the coolest race car driver you know?




The Stig.




Maybe make a "standard driver" that looks/weighs the same as the Stig.

Using this driver for comparison lap times would be a big equalizer.

JAT

[joevenzon]

Oooh, that's a sweet idea!

http://www.topgear.com/content/timetobur...8/1024.jpg

What do you think, zimluura?

Regarding the coordinate systems, when you say mass positioning, are you basing that off of the driver drawing position? Because the driver drawing position is probably different than the physics calculations based on the mass positions. I think that everything agrees except the driver drawing position (even if the driving drawing position makes more sense and agrees with blender). Is that your understanding as well?

[zimluura]

yeah, i'm basing my opinion on mass from how the driver adjustments are happening. 'cause his position *is* based on a mass position...

i assumed his visual reference model was being drawn where the mass is as far as the physical calculations are. but if they're not, i can't look into it too much without source diving. unless maybe i could try to make the driver weigh a few tons and then position him high and to one side to see which way the car tips...wow that sounds like a fun test to do.

<a>
so my 9000kg driver caused the car to fall through the track. probably stressed the suspension so much the car bottomed out in less than 1 frame and that took the collision box below the track...of course after that it was pretty much a mess.

<continued>
with a 400kg driver 2 meters up and 4 meters to the negative x the car seems to do a wheelie when accelerating forward. with a 200kg driver 4 meters positive x when i brake i get the back wheels into the air and the car does a forward somersault.

so i think x is maybe forward and back now, with negative being towards the back

when i place a 200kg driver 3 meters up and -4 meters of Y, the suspension on the right side seems compressed. and with positive 4 Y the suspension on the left compresses.

the nature of these tests does assume that positive z is up. which seems to jive with how the car handles with a 200kg driver 5 meters underground. and also how it's very easy to flip with a 200kg driver 5 meters up.

so my revised idea on how mass positioning currently works is this:
x = back and forward, negative being back
y = right and left, negative being right
z = down and up, negative being down


// edit
for the record, i'm in favor this:
x = left and right, negative being left
y = back and front, negative being back
z = down and up, negative being down

also if all the coordinate systems get all synchronized. then the driver-view position can be computed from his mass position + a vector for every vehicle. and that vector can be set in a .driver file...though i think it should probably get the extension .ini, or maybe .cfg, or .config, or some other something. mostly because on a file system if you see something with a .driver after it you're going to wonder what the hell driver model that it. solaris maybe?? ok i'm digressing to the point of useless rambling. i'll stop now.

[zimluura]

as for the Stig...

some say he's made up of polygons...
and if you were to look under his visor you'd just see the back of his helmet...


the picture is making me notice how my driver's proportions are more heroic and less realistic (smaller head/helmet, big upper body). good thing i can fix that without altering the UV map.

it'd be sorta trivial for me to make the Stig with most of the driver already modeled. it's mostly the helmet i'd be changing really. i'll try to modify my guy's proportions and then make a 2nd driver with a different helmet.

though - with small variations like that i'm tempted to request something where you pick your body, then your head and then your hairstyle/helmet, glasses/visor, threads/tatoos. but thats maybe better for the eventual skeletal system that is hopefully a very very low priority while the coordinate systems get straightened out.